BoilerLab is an international company with a profitable and sustainable business. We are actively expanding our product line and developing our ecosystem in three areas simultaneously: AI & ML, SaaS, and blockchain.

One of our products is Atomic Chat: an open-source local AI application and inference engine. It runs open-weight models directly on the user's device, without the cloud. Under the hood, it uses llama.cpp and MLX, an OpenAI-compatible server on localhost.

We are looking for a C++ developer for this project.

What you will be doing:

— Developing our fork of llama.cpp and keeping it in sync with upstream. — Promptly adding support for new models and architectures as soon as they are released. — Quantization and KV-cache compression, inference acceleration. — Low-level optimization: memory, latency, speed. — Support for different backends and hardware. — GPU programming and performance profiling. — Support and development of an OpenAI-compatible inference server. — Monitoring the development of inference engines and quickly integrating the best solutions into our project.

What we expect:

— Solid C/C++ experience, with experience in large codebases. — Experience with llama.cpp or another inference engine. — GPU programming experience (CUDA, Metal). — Understanding of quantization. — Cross-platform build experience with CMake. — Ability to profile and optimize performance.

Will be a plus:

— Contributions to llama.cpp / ggml. — Experience with MLX. — Knowledge of Rust.

What we offer:

— Work in a growing international company. — Payments in rubles, currency, or USDT. — Individual growth plan. — Coverage of courses, seminars, and relevant literature. — 28 calendar days of paid vacation, sick leave — 100%. — Flexible start to the workday from 10 AM to 12 PM. — Office-based work in Saint Petersburg. — Open source and product influence. — Small team and minimal bureaucracy.